Hence, please, don't evaluate any new ML algorithm on MNIST (solely), ever
Show this thread
It's the results you get that make your algorithm interesting, not how elegant your theory is or how much time you spent developing it
-
-
-
Instead, spend as much time as possible demonstrating that you are getting good results across a range of tasks/datasets as wide as possible
Show this thread -
Since getting these results is the whole point of what you are doing -- not an afterthought
Show this thread
End of conversation
New conversation -
-
-
I disagree. Good results might be the goal, but ad hoc work is less likely to be adopted. Elegance is easier to incrementally improve.
-
This is the essence of the scientific method: empirical validation trumps elegance. Most elegant theories turn out to be wrong.
- Show replies
New conversation -
-
-
If you work in a company, otherwise the ideas behind it may be more important to move forward the field
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
I agree with the general sentiment. However in that specific case, strong arguments can be made why elegance indicates something deeper.
-
e.g. most NN work remains on equivariance. However, steerability is far more interesting, even there moving beyond regular representations.
- Show replies
New conversation -
-
-
This is very short-sighted from a scientific point of view.
-
It is a simple rephrasing of the scientific method. An idea matters if it gets empirically validated, not if its creator is very smart.
- Show replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.